Support triggering flush by wal size

[Rachelint]

Describe This Problem

I found wal size will become too large in some situations, that leads to the unacceptable long recovery time. Maybe we should support triggering flush by wal size, this is a common flush triggering strategy implemented in many other dbs(e.g. rocksdb).

Proposal

• Support a method like should_flush in WalMananger.
• Peridic checking this and trigger flush.

Additional Context

No response